This document summarizes a presentation about research assessment metrics. It discusses traditional metrics like the journal impact factor and h-index, as well as alternative article-level metrics and altmetrics that aim to better capture broader impact. While metrics can provide useful information, they should be interpreted cautiously and used as only part of research assessment. Declarations recommend not using journal impact factors to assess individual researchers or research articles. Emerging altmetrics attempt to measure impact beyond citations, but more evidence is still needed to understand what altmetrics indicate.
1. Research Assessment:
What you need to know.
A Presentation by Ian Gibson, Collections Librarian
James A. Gibson Library
2. It’s all his fault
Image from http://vw.indiana.edu/places-
images/authors/eugene-garfield.jpg
3. The Journal Impact Factor
•𝐼𝐹 =
𝐶𝑖𝑡𝑎𝑡𝑖𝑜𝑛𝑠
𝐶𝑖𝑡𝑎𝑏𝑙𝑒 𝐴𝑟𝑡𝑖𝑐𝑙𝑒𝑠
• Where
– Citations = citations in the current year to articles published in
the 2 previous years
– Citable Articles = the number of articles published in the 2
previous years (in theory)
4. if you use impact
factors you are
statistically
illiterate1
1Stephen Curry, Sick of Impact Factors
http://occamstypewriter.org/scurry/2012/08/13/sick-of-impact-
factors
Photo from http://www.theproteinwrangler.com/wp-content/uploads/2011/10/scurry.jpg
5. Alternatives to Impact Factors?
• 5 Year Impact Factor
• Eigenfactor
• Source-Normalized Impact per Paper (SNIP)
• SCImago Journal Rank (SJR)
• Google Journal Rank
6. What Metrics Exist to Judge
Individuals?
Let me hook you up with
my h-index
Jorge Hirsch
http://www-physics.ucsd.edu/fac_staff/fac_profile/faculty_description.php?person_id=99
7. h-Index
• h = n papers that have been
cited at least n times.
9. “Citation‐based statistics can play
a role in the assessment of
research, provided they are used
properly, interpreted with caution,
and make up only part of the
process. Citations provide information about journals,
papers, and people. We don't want to hide that information; we
want to illuminate it.”
- Joint Committee on Quantitative Assessment of Research, International Mathematics
Union in cooperation with the International Council of Industrial and Applied Mathematics
(ICIAM) and the Institute of Mathematical Statistics (IMS)
http://www.mathunion.org/fileadmin/IMU/Report/CitationStatistics.pdf
10. Declaration on Research Assessment
General Recommendation
1. Do not use journal-based metrics, such
as Journal Impact Factors, as a surrogate
measure of the quality of individual research
articles, to assess an individual scientist's
contributions, or in hiring, promotion, or
funding decisions.
http://am.ascb.org/dora/
11. Altmetrics
• http://altmetrics.org/manifesto
• Putting the focus at the article level
• Capturing broader impact:
– “Hallway chats” (e.g. twitter, blog postings)
– Those papers that you don’t cite but keep in that
drawer in your office (e.g. saved in a reference
manager like Zotero or Mendeley)
– The papers that you read just for general interest
(e.g. download statistics)
19. Bibliography
• Brembs, B., Button, K., & Munafò, M. (2013). Deep impact: unintended consequences of journal rank. Frontiers in Human Neuroscience, 7, 291.
doi:10.3389/fnhum.2013.00291
• Cantrill, S. (2012, June 30). Inside our Impact Factor. The Sceptical Chymist. Retrieved
fromhttp://blogs.nature.com/thescepticalchymist/2012/06/inside-our-impact-factor.html
• Colquhoun, D., & Plested, A. (2014, January 16). Why you should ignore altmetrics and other bibliometric nightmares. DC’s Improbable Science.
Retrieved April 2, 2014, from http://www.dcscience.net/?p=6369
• Costas, R., Zahedi, Z., & Wouters, P. F. (2014, January 17). Do “altmetrics” correlate with citations? Extensive comparison of altmetric indicators with
citations from a multidisciplinary perspective. CWTS-WP-2014-001. External research report. Retrieved April 2, 2014,
fromhttps://openaccess.leidenuniv.nl/handle/1887/23041
• Curry, S. (2012, August 13). Sick of Impact Factors. Reciprocal Space. Retrieved April 7, 2014,
from http://occamstypewriter.org/scurry/2012/08/13/sick-of-impact-factors/
• Fenner, M. (2013). What Can Article-Level Metrics Do for You? PLoS Biol,11(10), e1001687. doi:10.1371/journal.pbio.1001687
• Garfield, E. (1955). Citation Indexes for Science. Science, 122(3159), 108–111. Retrieved from http://www.jstor.org/stable/1749965
• Garfield, E. (1964). “Science Citation Index”-A New Dimension in Indexing.Science, 144(3619), 649–654. Retrieved
fromhttp://www.jstor.org/stable/1712875
• Garfield, E. (2005). The Agony and the Ecstasy— The History and Meaning of the Journal Impact Factor. Presented at the International Congress on
Peer Review And Biomedical Publication, Chicago. Retrieved fromhttp://garfield.library.upenn.edu/papers/jifchicago2005.pdf
• Rossner, M., Epps, H. V., & Hill, E. (2007). Show me the data. The Journal of Cell Biology, 179(6), 1091–1092. doi:10.1083/jcb.200711140
• Seglen, P. O. (1992). The skewness of science. Journal of the American Society for Information Science, 43(9), 628–638. doi:10.1002/(SICI)1097-
4571(199210)43:9<628::AID-ASI5>3.0.CO;2-0
• Sud, P., & Thelwall, M. (2014). Evaluating altmetrics. Scientometrics, 98(2), 1131–1143. doi:10.1007/s11192-013-1117-2
• The PLoS Medicine Editors. (2006). The Impact Factor Game. PLoS Medicine, 3(6), e291. doi:10.1371/journal.pmed.0030291
• Thelwall, M., Haustein, S., Lariviere, V., & Sugimoto, C. R. (2013). Do Altmetrics Work? Twitter and Ten Other Social Web Services. PLoS ONE, 8(5),
e64841. doi:10.1371/journal.pone.0064841
• Vanclay, J. K. (2012). Impact factor: outdated artefact or stepping-stone to journal certification? Scientometrics, 92(2), 211–238. doi:10.1007/s11192-
011-0561-0
Editor's Notes
This presentation is about awareness – research assessment is not a topic that most of you will have reason to encounter in your day to day life.
This is Eugene and it’s all his fault
In 1955 my boy here decided that “science” was too difficult to follow given the limitations of subject indexing. What was needed was a means of easily following ideas. Law provided the example of a functioning citation based indexing system in Sheppard’s Citations. In 1961 the Genetics Citation Index was launched. This eventually grew into Science Citation Index and into the Web of Science
Most importantly it made counting citations easy and desirable; small step in the quantification of quality
Introduced a few issues
Transcription errors actually make counting citations much more difficult
Which citations actually count? There is no comprehensive citation index – not an issue if we don’t care about new journals; patents/grey literature; book coverage is spotty at best (though TR and Google are trying to rectify this)
Context of citation – citation does not imply quality; citation does not imply agreement; citation does not imply importance
It became obvious rather quickly that a number of metrics could be created using these citation counts to judge quickly the importance of articles, journals, and people. Garfield initially mused in 1955 that one could create an “impact factor” to judge the impact of individual articles. Both the initial Genetics Citation Index concept and the larger Science Citation Index were based around author… however if you switch the focus from author to journal title…
Initially Garfield and Sher developed the JIF to determine what journals to include in the Science Citation Index – no idea that it would become such a controversial figure (see http://garfield.library.upenn.edu/papers/jifchicago2005.pdf)
When the numbers were released publicly in the 70s the idea was that it was a tool for libraries to help decide how to allocate our journal subscriptions (i.e. which titles formed a “core” collection)
The formula looks straight forward enough but when you start digging into it one sees it as a deeply flawed metric
For instance what citations are we counting? Well all the citations obviously except that we’re not really – it’s all the citations coming from WoS indexed content – that list is getting more comprehensive with the inclusion of books, conferences and data recently but we’re still not getting patents, gov’t reports, grey literature, preprints, etc. But there isn’t any distinction between citations to news, opinion, and review pieces and cites of research articles – doesn’t seem like a big deal but…
What counts as a citable article? PLOS Medicine published an article detailing the negotiations around what would count as citable when they go their first impact factor (http://www.plosmedicine.org/article/info:doi/10.1371/journal.pmed.0030291). Journals have an obvious interest in keeping the denominator low and the numerator high
For an interesting look at what happens when you manipulate see Stuart Cantrill’s analysis of how publishing review aritcles inflated the impact factor of Nature Chemistry from 16 up to 20.
There are also questions about reproducibility – Rockefeller University Press tried to reproduce their impact factors and the IFs of their competitors first using just WoS then using data provided by TR – despite multiple attempts they could not find the magic numbers to make the calculations work (see http://jcb.rupress.org/content/179/6/1091.full)
This is Stephen Curry a BioChemist @ Imperial College London he wrote a blog post talking about being sick of the use of impact factors as some kind of signifier of quality for individuals.
This post mentioned most of the major critiques of IF
Citation distribution is skewed Per Seglens showed in 1992 that 15% of the articles in a typical journal are responsible for ½ total citations – so journals are being rewarded for publishing a few good papers in a sea of mediocrity
How it is being used in grant aps/demanded by granting bodies even though it says nothing about the individuals actual contributions to that particular journal
Highlights a paper by Vanclay (http://arxiv.org/abs/1201.3076) that points out issues around a short time frame; doesn’t account for different citation practices; is selective, etc.
One of the interesting consequences of this calculation is that IF should increase every year – for the simple reason that every year more and more stuff is being published and therefore more and more citations
So we know that IF isn’t great what are the alternatives
Eigenfactor measures the influence of journals using a method similar to Google’s PageRank algorithm – the inventors claim that by using a longer timeframe 5 years and their enhanced algorithm they can eliminate many of the problems of the JIF – Basis in Web of Science but with additional data put out by University of Washington Researchers
SNIP based on Scopus data + 3 previous years – aims to correct for variations in citing behaviour between disciplines – claims better representation b/c Scopus indexes more journals (though there is some debate about how comprehensively Scopus actually covers those journals…) put out by CWTS @ U Liedan
SJR also based on Scopus data – uses the same ideas as Eigenfactor/Page Rank – citations weighted by the quality of citing source – again weighting is used to correct for differing citation practices between fields
Google Journal Rank – based on the H-Index – the h5 index calculates the h-index based on a 5 year window
THESE ARE ALL JOURNAL BASED METRICS
Brembs, Button & Munafo (http://journal.frontiersin.org/Journal/10.3389/fnhum.2013.00291/full) suggest that journal rank (and in fact journals themselves) need to be done away
(1) journal rank is a weak to moderate predictor of utility and perceived importance;
(2) journal rank is a moderate to strong predictor of both intentional and unintentional scientific unreliability;
(3) journal rank is expensive, delays science and frustrates researchers; and,
(4) journal rank as established by IF violates even the most basic scientific standards, but predicts subjective judgments of journal quality
Anyone see a problem with this?
It rewards a long career or a prolific career
It doesn’t reward pathbreaking ideas and papers that get a lot of citations
It doesn’t account for multiple authors
Rewards prolific authors (e.g. if you’ve only published 1 paper the highest your h-index could be is 1)
Regardless of whether you feel we actually can We have a proliferation of metrics all being built on the same creeky foundationg
Citation based metrics have started to creep into disciplines well outside the scientific core (e.g. Business and education)
- 2008 report on citation statistics found that citation statics though “easy” were not well understood “Numbers are not inherently superior to sound judgement”
-the declaration on research assessment was formulated by the American Society for Cell Biology (in conjunction with others) to shed light on the wide spread practice of judging grant applications and T&P decisions on where the applicants had published instead of what they published.
Altmetrics sprang from a parallel movement to offer some kind of article level metrics or at least article level data on citations & downloads.
Altmetrics is trying to capture impact that research has in and out of academia
Research is broadly defined (right now anything that you can give a permanent identifier to… blog posts, git hub repositories, etc.
Trying to capture not just use but also some of the nuance behind the use.
The important point is that a lot of interaction with an article has to happen before it is cited and previously we couldn’t capture much of that process
PLOS was one of the first publishers to give metrics publically displayed for each article
Useful categorization: Viewed, Cited, Saved, Discussed, and Recommended
“All PLOS Biology articles are viewed and downloaded, and almost all of them (all research articles and nearly all front matter) will be cited sooner or later. Almost all of them will also be bookmarked in online reference managers, such as Mendeley, but the percentage of articles that are discussed online is much smaller. Some of these percentages are time dependent; the use of social media discussion platforms, such as Twitter and Facebook for example, has increased in recent years (93% of PLOS Biology research articles published since June 2012 have been discussed on Twitter, and 63% mentioned on Facebook). These are the locations where most of the online discussion around published articles currently seems to take place; the percentage of papers with comments on the PLOS website or that have science blog posts written about them is much smaller.”
What Can Article-Level Metrics Do for You? http://dx.doi.org/10.1371/journal.pbio.1001687
PLOS has analyzed this data and come up with some interesting things:
HTML views seem to be driven mostly by social sharing
PDF views have no correlation to social sharing
Most articles have 4:1 HTML to PDF views but as that ratio increases there is a very strong correlation to social sharing.
Suggests that social sharing is a good indicator of public engagement
They don’t see much correlation between ppl saving in a citation manager (like Mendeley) & citation counts
They find that meaningful citation numbers don’t accumulate for 2-5 years even with the typically high citation culture of the life sciences
Impact story is a not for profit open source project that seeks to contextualize the raw numbers that you might find at PLOS
Right now only limited data is available.
Can input DOIs, URLs, github username + repository name or use your ORCID
Show final report in link then show input screen.
Can use your ORCID 0000-0001-9154-8247
Similar to Impact story but right now mostly focused on NPG products
Can use their free bookmarklet which will automatically provide a report
Altmetrics are quite new – so it’s still under development
Working paper by Costas, Zahedi and Wouters from University of Leiden showed that many articles have very little altmetric activity H&SS along w/ biomed has the most. Much research still to be done
Thelwell et al. (http://dx.doi.org/10.1371/journal.pone.0064841) in PLOSOne report that significant altmetric activity (for certain services) does correlate to higher citation but that use of some altmetric services are so seldome used it might not actually be worthwhile
Sud and Thelwell (http://dx.doi.org/10.1007/s11192-013-1117-2) recommend methods for evaluation altmetrics
David Colquhoun writes on his blog that it’s best to ignore altmetrics and other bibliometric nightmares “Altmetrics are numbers generated by people who don’t understand research, for people who don’t understand research. People who read papers and understand research just don’t need them and should shun them.”
Despite what I’ve said today I’m not below keeping tabs on my paper
Other key takeaways –
Citations are difficult to deal with and any measure built on them must be taken with a grain of salt.
Impact Factors are not the be all and end all and are in fact highly flawed; they have their place but you don’t have one.
Many bibliometric scores are at the journal level